Repository navigation
terminal-bench: Wait for Docker and make background-shell require a background task - #1255
Open
max-rozen-oss-db wants to merge 2 commits into
Open
max-rozen-oss-db wants to merge 2 commits into
max-rozen-oss-db wants to merge 2 commits into
Conversation
…ground Harbor's preflight gives `docker info` 10 seconds, which failed the TB2 step on a fresh runner. Wait up to two minutes for the daemon first. When Smart Router picks Opus for background-shell, Claude backgrounds the job and ends its turn, and `claude -p` kills the job. Disable background tasks for that task so Claude waits in the foreground, as Haiku already does. Co-authored-by: Isaac <no-reply@databricks.com>
max-rozen-oss-db
requested review from
AarushiShah-db,
lilly-luo and
rohita5l
as code owners
October 22, 2026 21:35
The job now prints a token and waits for the agent to ack it in ack.txt, so a foreground call can never finish it. The verifier also checks that Claude started the job with run_in_background. Drops the earlier change that disabled background tasks for this task. Co-authored-by: Isaac <no-reply@databricks.com>
max-rozen-oss-db
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes two flaky Terminal Bench failures.
docker info10s and failed with "Docker daemon is not running". The same runner reached Docker fine 13 minutes later. A new step waits up to 120s fordocker infobefore harbor runs.background-shellnever required a background task. Haiku ran the job in the foreground and always passed. Opus backgrounded it, ended its turn, andclaude -pkilled the job. Opus failed all 3 of its attempts across the last 7 runs.background-shellnow tests background tasks directly:slow_job.pyprints a token and waits for the agent to write it toack.txt, then runs 90s more and prints the result. A foreground call can't finish it.run_in_backgroundcall that startedslow_job.py. Codex logs aren't stream-json, so Codex is covered by the ack alone.Expect the Claude lanes to fail this task if a model ends its turn while the job runs.
Testing
run_native.py(91s).This pull request and its description were written by the Isaac LLM tool (Claude Code, session
276df331-02af-4bc3-84a4-5e458f178db2).